Representative photo: Mufid Majnun/Unsplash
Bengaluru: Two recent studies on antibody levels among Covishield and Covaxin recipients sparked a great deal of controversy because they made the Serum Institute of India jab look better than that of Bharat Biotech.
The first study, a pan-India survey of healthcare workers, reported that people who got Covishield made more of a type of antibody – called the spike antibody – to the SARS-COV-2 virus, relative to Covaxin recipients. Moreover, 20% of Covaxin recipients didn’t have detectable spike antibodies a month later, while the number was only around 2% for Covishield.
The second study, a smaller one of rheumatic disease patients, echoed these findings.
Unsurprisingly, the studies fuelled news stories that Covishield had a “higher efficacy” – a measure of how well the vaccine protects against disease – than Covaxin. Some outlets were more cautious in their interpretation, refusing to draw efficacy comparisons while reporting the antibody differences between the jabs.
Meanwhile, even though the studies’ authors hadn’t speculated on efficacy in their manuscripts, they did express worry, on Twitter, about the low antibody titres after Covaxin. One of them wrote that the complete lack of spike antibodies in over 20% of Covaxin recipients needed “urgent introspection”.
The unfavourable comparison with Covishield provoked a prickly response from Covaxin’s maker Bharat Biotech. Raches Ella, the firm’s project lead for COVID-19 vaccines, slammed the first study on Twitter. Claiming that the spike antibody titres couldn’t predict efficacy, he accused the authors of publicising a study that hadn’t yet been peer-reviewed.
(The studies had been posted through preprint papers.)
Which of the many study interpretations thrown around was accurate? And can’t we predict vaccine efficacy from antibody response?
The Wire Science spoke to several scientists and reviewed the literature to find out. The answer is, as with all things immunology, complicated.
What exactly did the studies find?
The first study, by a group of doctors led by Awadhesh Kumar Singh, a Kolkata-based endocrinologist, compared 425 healthcare workers who received two Covishield doses with 90 Covaxin recipients. To do this, they used a test called Liaison, manufactured by Italian firm DiaSorin, that detects antibodies to the virus’s spike protein.
Singh and his team found that 98% of Covishield recipients became seropositive – their blood had detectable levels of antibodies – around one month after the second dose; 80% of Covaxin recipients did the same. Further, the median spike antibody titres (levels) were higher in the Covishield arm than in the Covaxin arm.
The second study, led by Kochi-based rheumatologist Padmanabha Shenoy, compared serum samples from 102 vaccine recipients with autoimmune rheumatic disease, and 34 without. They used a different test for spike antibodies, called the Elecsys Anti SARS-COV-2 S assay, manufactured by Swiss firm Roche.
Their findings: 95% of the 120 patients who received two Covishield doses and 68.7% of the 16 who received two Covaxin doses had detectable levels of antibodies. Again, Covishield induced a higher antibody titre than Covaxin.
This trend of higher antibody titres and seropositivity with Covishield persisted when they repeated the study with 30 healthy recipients of each jab. All Covishield recipients became seropositive after two jabs – while only 23 (76.66%) of Covaxin recipients did.
What are spike antibodies?
They are one of the many antibodies that humans make when they encounter either the SARS-CoV-2 virus or vaccines against the virus.
SARS-CoV-2 is a spherical virus made up of several proteins. Four of these proteins make up the virus’s structure: the membrane protein and envelope protein build the viral coat; the nucleocapsid protein protects the coiled-up viral genome inside; and the spike protein gives the virus its characteristic look of a ball with barbs all over it.
The spike protein also acts like a key that allows the virus to invade a human cell. The keyhole in this case is an enzyme called ACE2 on the surface of human cells, whose structure allows the spike to ‘unlock’ the cell.
People fight a SARS-CoV-2 infection by producing a range of antibodies, cells and other proteins. Scientists refer to the antibody response as humoral immunity; immune cells (the so-called T-cells, macrophages, etc.) are collected under cellular immunity.
Both humoral and cellular immunity target the full range of proteins in the SARS-CoV-2 virus, including the spike proteins.
And of this lot, the antibodies that target the spike proteins are called spike antibodies.
If immunity is so wide-ranging, why did both studies measure only spike antibodies?
Out of the many bits that make up the humoral and cellular immunities, one type of antibody – called a neutralising antibody – plays an outsize role in preventing human cells from getting infected.
Several studies have shown that the vast majority of these neutralising antibodies are also spike antibodies. In fact, according to one study, over 90% of neutralising antibodies in the bodies of convalescent people – i.e. people who have recovered from COVID-19 – target only one portion of the spike protein, called the receptor binding domain. By doing this, neutralising antibodies disable the ‘key’ that is the spike, so that the virus can’t ‘unlock’ a human cell anymore.
There is also some evidence that the neutralising antibody titres after vaccination could predict a vaccine’s efficacy – although this is far from certain. We need more research to establish exactly what neutralising titre is correlated with what efficacy level.
(See question: Why don’t we know yet what neutralising titres can predict efficacy?)
Given how many other parts of the immune system are involved in determining a vaccine’s efficacy, it’s possible we may never find such a neat association, called an ‘immune correlate of protection’. Ultimately, we may find that neutralising antibodies alone can’t predict efficacy, and that we will need to measure cellular immunity as well.
Still, neutralising antibodies are promising, which is why scientists often measure them to get a broad idea of whether a vaccine is working.
Singh and Shenoy, the lead authors of the two studies, wanted to compare the performance of Covaxin and Covishield. And they also initially considered measuring neutralising titres after vaccination.
However, doing this requires tests called the plaque reduction neutralisation test (PRNT) or the microneutralisation assay (MN). Both of them are expensive, hard to perform and not widely accessible in India.
So, in the absence of PRNT and MN tests, both groups measured the next best thing: spike antibodies. Because neutralising antibodies are mostly spike antibodies (the first is a subset of the second), the levels of both tend to track each other.
Some studies have shown a correlation between spike antibody titres and neutralising titres. But this correlation isn’t perfect. A small fraction of people with spike antibodies don’t have neutralising antibodies, and vice versa.
This is where the caveats in interpreting studies like those by Singh and Shenoy come in: spike antibodies correlate imperfectly with neutralising antibodies, which may or may not correlate with efficacy. This tenuous relationship means we must be cautious in claiming a direct link between spike antibody titres and efficacy.
Why don’t we know what neutralising titre can predict efficacy?
Because everyone uses different tests to measure neutralising antibodies.
For example, in the phase 2 trial for its mRNA vaccine, firm Pfizer measured neutralising antibodies using a PRNT test designated IC50. The ’50’ means, broadly, that the test measures the maximum number of times the antibody-containing serum of a vaccinated individual can be diluted before it kills 50% of SARS-CoV-2 virus particles. The results are typically expressed in a ratio: if the serum was diluted 80 times before it stopped neutralising 50% of the virus, the neutralising antibody titre is 1:80.
Now, if researchers everywhere on the planet used the PRNT IC50 assay, they could assume that the neutralising titres in the Pfizer vaccine’s phase 2 trial were correlated with the 95% efficacy found in the vaccine’s phase 3 trial.
But this is not the case. Every vaccine manufacturer has been using a different neutralising test. In its phase 2 trial, Moderna used the PRNT 80, Bharat Biotech used the PRNT 50, AstraZeneca used the MN IC 50 and Novavax used the MN IC>991.
Each of these tests is as different as its name suggests: they all measure different levels of viral reduction, using the same broad principles, but different materials and methods. As a result, we can’t compare the results of these tests, said Gagandeep Kang, a public health microbiologist at the Christian Medical College, Vellore.
In other words, a neutralising titer of 1:80 in a PRNT IC50 is different from the same titre in an MN IC 50. What’s more, even the results of a PRNT 50 done in one lab could differ from a PRNT 50 in another, Kang added.
As a result, despite a mountain of data on neutralising titres in vaccine recipients, we still can’t say if there is a common titre that predicts a certain level of efficacy.
Will neutralising titres ever give us an immune correlate of protection then?
It’s possible – because immune correlates of protection do exist for many older vaccines.
Some vaccines, like those of tetanus, diphtheria and measles, have very strong correlates, called ‘absolute correlates’. If a person who got a measles jab showed a neutralising antibody titre of 120 mIU per ml (mIU is an international unit for measles antibody levels), for example, they are almost always protected against disease.
“If you develop such an absolute correlate, you’d be in a happy situation, where you wouldn’t need any efficacy study for vaccines,” Kang said. “And you could do all sorts of comparisons between vaccines using antibody levels.”
Other vaccines, like those for influenza, have relative correlates of protection. This means above a certain antibody titre, protection is likely but not guaranteed.
Developing such correlates takes many years of work, according to Kang. And for COVID-19, this work has just begun.
Earlier this year, a group of Australian scientists took an early first step in this direction. To get around the differences between neutralising tests used in vaccine trials, they took advantage of the fact that these trials also measure neutralising titres in convalescent people using the same tests they used with vaccine recipients.
Because convalescent neutralising titres are likely to be similar across trials, the scientists reasoned that they could express post-vaccination neutralising titres in terms of the convalescent titres. Finally, because post-vaccination neutralising titres could be mapped to efficacy in phase 3 trials, they could now map convalescent neutralising titres to efficacy as well.
Their study found that across seven vaccine trials, a 50% efficacy level correlated well with 20% of the mean neutralising titre among convalescent people. This level, in turn, mapped to neutralising titres of 1:10 to 1:30 on the widely different tests used in the trials.
But does this mean that 20% of the mean neutralising titre in convalescent individuals always results in 50% efficacy? In other words, do we have the elusive immune correlate of protection for COVID-19?
Far from it, said Satyajit Rath, an immunologist at the Indian Institute of Scientific Education and Research, Pune. Because the Australian group’s paper made several assumptions, including that all convalescent people had similar levels of antibodies, its results must be interpreted carefully.
“The paper makes a lot of approximations, and is built on a very rickety structure, which it acknowledges,” Rath said.
Another way to estimate an immune correlate of protection would be to translate the results of one neutralising test to another. This could happen if all neutralising tests in the world tested the same serum sample having a known antibody titre. Scientists have carried out such exercises in the past. For example, the UK government’s National Institute of Biological Standards and Control (NIBSC) sells a serum sample containing measles antibodies that test makers can use to standardise their kits.
Last year, the WHO began work on a similar international standard for COVID-19 antibody tests. Scientists can now purchase this standard – a sample of convalescent plasma collected from COVID-19 patients – from NIBSC to standardise their neutralising antibody kits.
As more and more labs express their results in terms of the WHO standard, Kang said, it may become easier to arrive at an immune correlate for COVID-19.
So we don’t have a neutralising antibody correlate for efficacy. But the two studies found over 20% of Covaxin recipients to be seronegative for spike antibodies. Isn’t that a bad thing?
No – because seronegativity in both studies doesn’t mean there are no spike antibodies. It only means that the tests the authors used to measure antibodies weren’t sensitive enough to detect antibodies below the seronegativity cutoffs.
The Liaison test for spike antibodies that Singh & co. used, for example, has a seronegativity cutoff of 15 AU/ml 2. However, because we don’t know how spike antibody levels are correlated with neutralising antibodies, and how neutralising antibodies are correlated with efficacy, even the few spike antibodies below the 15-AU/ml level could be enough for substantial vaccine efficacy.
In fact, the manufacturer of the Liaison test notes that, according to studies it performed, the 15-AU/ml level correlated with a 1:40 titre on a PRNT 90 neutralising assay. Rath pointed out that this level is higher than the neutralising titres (range of 1:10 to 1:30) mapped to 50% efficacy in the Australian study.
This simplistic comparison suggests that the seronegativity cutoff for the Liaison test is too high to mean much for protection. More importantly, it shows why we can’t compare results from different studies.
Finally, the human immune response to SARS-CoV-2 may be too complex to be reduced to spike-antibody titres alone, according to Rath. Covaxin likely also triggers cellular immunity, and neither the Singh et al nor Shenoy et al studies measured this. Most scientists trying to characterise the immune response to Covid today tend to measure only antibody responses, and not cellular immunity, because the first kind of test is easier to do.
But this doesn’t give the full picture. Relying on antibody tests alone, because it’s easier, is “a bit like losing your key somewhere, and finding your nearest streetlight and looking under that streetlight, because that’s the easiest thing you can do,” Rath said.
So how do we know Covaxin’s efficacy?
Only through a phase 3 clinical trial.
Unfortunately, six months since Covaxin was licensed for use in India, Bharat Biotech has still not published the full results from this trial for the vaccine. Indeed, it is to fill this gap in information about Covaxin that Singh and Shenoy conducted their studies in the first place.
This in turn makes Raches Ella’s criticism of Singh’s study on Twitter both hypocritical and unfair.
Hypocritical because Ella accused the scientists of extrapolating efficacy based on spike antibody titres – whereas his own company sought regulatory approval for its vaccine based on neutralising antibody titres and T-cell response. Neither is a particularly accurate predictor of efficacy.
And unfair because – despite the many difficulties in interpreting Singh’s and Shenoy’s data – both studies were important. They represented the first instance of scientists comparing Covaxin and Covishield using the same spike-antibody tests. Given the lack of comparability between various antibody tests, then, their findings were significant, even if they weren’t the last word on both the vaccines.
The reporting for this article was supported by a grant from the Thakur Family Foundation. The foundation did not exercise any editorial control over the contents of the article.
Priyanka Pulla is a science writer.